bioRxiv preprint first posted online Mar. 20, 2017; doi: http://dx.doi.org/10.1101/118372. The copyright holder for this preprint (which was not
peer-reviewed) is the author/funder. It is made available under a CC-BY 4.0 International license.
Genomic inferences of domestication events are corroborated by written records
in ​Brassica rapa
Xinshuai Qi​1,7​, Hong An​2,3,7​, Aaron Ragsdale​4​, Tara E. Hall​1​, Ryan N. Gutenkunst​5​, J. Chris
Pires​2​, Michael S. Barker​1,6
1. Department of Ecology & Evolutionary Biology, University of Arizona, Tucson, Arizona, USA
85721
2. Division of Biological Sciences, University of Missouri, Columbia, Missouri, USA 65211
3. National Key Lab of Crop Genetic Improvement, Huazhong Agricultural University, Wuhan, P.
R. China 430070
4. Program in Applied Mathematics, University of Arizona, Tucson, Arizona, USA 85721
5. Department of Molecular and Cellular Biology, University of Arizona, Tucson, Arizona, USA
85721
6. Author for correspondence: ​[email protected]
7. These authors contributed equally to this work
Key words:
Brassica rapa​, genetic structure, transcriptome, domestication, crop improvement, population
genomics.
1
bioRxiv preprint first posted online Mar. 20, 2017; doi: http://dx.doi.org/10.1101/118372. The copyright holder for this preprint (which was not
peer-reviewed) is the author/funder. It is made available under a CC-BY 4.0 International license.
Abstract
Demographic modeling is often used with population genomic data to infer the
relationships and ages among populations. However, relatively few analyses are able to validate
these inferences with independent data. Here, we leverage written records that describe distinct
Brassica rapa​ crops to corroborate demographic models of domestication. ​Brassica rapa​ crops
are renowned for their outstanding morphological diversity, but the relationships and order of
domestication remains unclear. We generated genome-wide SNPs from 126 accessions
collected globally using high-throughput transcriptome data. Analyses of more than 31,000 SNPs
across the ​B. rapa​ genome revealed evidence for five distinct genetic groups and supported a
European-Central Asian origin of ​B. rapa​ crops. Our results supported the traditionally
recognized South Asian and East Asian ​B. rapa​ groups with evidence that pak choi, Chinese
cabbage, and yellow sarson are likely monophyletic groups. In contrast, the oil-type B
​ . rapa
subsp. ​oleifera​ and brown sarson were polyphyletic. We also found no evidence to support the
contention that rapini is the wild type or the earliest domesticated subspecies of ​B. rapa​.
Demographic analyses suggested that ​B. rapa​ was introduced to Asia 2400–4100 years ago,
and that Chinese cabbage originated 1200-2100 years ago via admixture of pak choi and
European-Central Asian ​B. rapa​. We also inferred significantly different levels of founder effect
among the ​B. rapa​ subspecies. Written records from antiquity that document these crops are
consistent with these inferences. The concordance between our age estimates of domestication
events with historical records provides unique support for our demographic inferences.
2
bioRxiv preprint first posted online Mar. 20, 2017; doi: http://dx.doi.org/10.1101/118372. The copyright holder for this preprint (which was not
peer-reviewed) is the author/funder. It is made available under a CC-BY 4.0 International license.
Introduction
Demographic analyses are widely used to infer the history of populations. Many analyses
make demographic inferences of ancient events that occurred in prehistory. Thus, there is often
no opportunity to validate these inferences with independent data. Domestication offers a unique
chance to verify the inferences from methods that are used in model and non-model species.
Artificial selection and domestication have long served as a testing ground for understanding
evolution ​(Darwin 1859; Diamond 2002)​. The study of crop domestication provides fundamental
insights on crop history and improvement, but also contributes some of our most well understood
systems to reveal the powers and limits of selection to shape organisms ​(Ross-Ibarra ​et al.
2007; Purugganan & Fuller 2009; Gross & Olsen 2010; Olsen & Wendel 2013; Meyer &
Purugganan 2013)​. ​Brassica​ crops are arguably one of the best examples to illustrate the
diversity that can be created by artificial selection ​(Purugganan ​et al.​ 2000; Prakash ​et al.​ 2011)​.
The three ​Brassica​ crops most well known for their morphological diversity are​ B. rapa​ (AA, 2n =
20), ​B. oleracea ​(CC, 2n = 18), and their allopolyploid derivative, ​B. napus​ (AACC, 2n = 38).
Each of these taxa includes several cultivated subspecies that were domesticated for a variety of
uses. Long considered a classic textbook example of plant domestication ​(Darwin 1859;
Ladizinsky 2012; Zohary ​et al.​ 2012)​ and the power of artificial selection ​(Purugganan ​et al.
2000)​, ​B. oleracea​ includes a diversity of crops such as broccoli, cauliflower, cabbage, kale,
Brussels sprout, and kohlrabi. However, its sister species, ​B. rapa​, has similar morphological
variation and represents a parallel example for comparative analyses of artificial selection and
evolution. Several subspecies are recognized in ​B. rapa ​that have been cultivated for particular
phenotypic characteristics. These include turnip (subsp. ​rapa​) with an enlarged edible root,
3
bioRxiv preprint first posted online Mar. 20, 2017; doi: http://dx.doi.org/10.1101/118372. The copyright holder for this preprint (which was not
peer-reviewed) is the author/funder. It is made available under a CC-BY 4.0 International license.
yellow (subsp. ​trilocularis​) and brown sarsons (subsp. ​dichotoma​) that are used as mustard and
oil seeds mainly in India, the oilseed field mustard (subsp. ​oleifera​), and the leafy East Asian
vegetables pak choi (subsp. ​chinensis​) and Chinese cabbage (subsp. ​pekinensis​) with their
enlarged mid-rib and leaves (Fig. 1). There are also several less broadly known leafy
subspecies, including rapini (subsp.​ sylvestris​), choy sum (subsp. ​parachinensis​), zi cai tai
(subsp. ​purpuraria​), tatsoi (subsp. ​narinosa​), mizuna (subsp. ​nipposinica​), and komatsuna (or
Japanese mustard spinach/rapini, subsp. ​perviridis​).
Domestication of the diverse ​B. rapa​ crops most likely occurred during written history
(after ~3500 BCE). Many of the classically studied crops, such as maize and Asian rice, were
domesticated in prehistory ​(Smith 1997; Piperno & Flannery 2001; Long ​et al.​ 2006; Kovach ​et
al.​ 2007; Piperno ​et al.​ 2009; Fuller ​et al.​ 2009; Molina ​et al.​ 2011; Sang & Ge 2013; Gross &
Zhao 2014)​. A written record of domestication for most crops is simply unavailable to develop
hypotheses and corroborate genomic analyses. In contrast, the oldest mention of ​Brassica​ crops
in general dates back nearly 5000 years.​ Brassica​ crops were first described in a Chinese
almanac from ~3000 BCE and ancient Indian texts from around 1500 BCE ​(Prakash ​et al.​ 2011)​.
Although the record is unclear on which European ​B. rapa​ crop was first domesticated, turnips
were on the list of plants grown in the garden of Merodach-baladan in Babylonia (722–711 BCE)
(Körber-Grohne 1987)​. Consistent with this early written record, archaeological evidence of
turnip seeds and roots have been reported from multiple archeological sites of Neolithic villages
in Switzerland, India, and China (Prakash and Hinata 1980; Ignatov et al. 2008). ​Brassica rapa
subsp. ​rapa​ was clearly described nearly 2600 years ago in Shi Jing, the oldest existing Chinese
classic poetry collection ​(Li 1981; Ye 1989; Luo 1992)​. The authors of Shi Jing used the ancient
Chinese character for turnip, indicating that this crop was in East Asia by that time. The earliest
4
bioRxiv preprint first posted online Mar. 20, 2017; doi: http://dx.doi.org/10.1101/118372. The copyright holder for this preprint (which was not
peer-reviewed) is the author/funder. It is made available under a CC-BY 4.0 International license.
historical records of Chinese cabbage (​B. rapa​ subs. ​pekinensis)​, “Niu Du Song (beef tripe
cabbage)”, appears in Xin Xiu Ben Cao, one of the earliest official pharmacy books edited by
Jing Su (659 CE). A clear description of Chinese cabbage is also found in Ben Cao Tu Jing
(1061 CE) by Song Su. He described Niu Du Song in Yangzhou city in Southern China as a
plant with large, round, and tender leaves with a uniquely wrinkled leaf. Based on these
documents, we may expect that many of the the Asian ​B. rapa​ crops were domesticated as long
as 3500 years ago and that​ B. rapa​ subsp. ​pekinensis​ may be at least 1500 years old.
Despite the economic importance and historical records of ​B. rapa ​crops, our
understanding of their relationships and origins is still not well resolved. Previous studies have
attempted to identify genetically discrete groups among ​B. rapa​ crops with molecular markers
(McGrath & Quiros 1992; Ren ​et al.​ 1995; Zhao ​et al.​ 2005, 2007, 2010; Takuno ​et al.​ 2006; Del
Carpio ​et al.​ 2011a; b; Guo ​et al.​ 2014; Pang ​et al.​ 2015; Takahashi ​et al.​ 2015; Tanhuanpää ​et
al.​ 2016)​. Most analyses identify three ambiguous geographic groups present in Europe and
Central Asia, South Asia, and East Asia ​(Del Carpio ​et al.​ 2011a; b; Guo ​et al.​ 2014; Pang ​et al.
2015)​ except a recent analysis that found no evidence for geographic structure among diverse
B. rapa​ accessions ​(Tanhuanpää ​et al.​ 2016)​. In addition, previous genetic and empirical studies
suggested two distinct hypotheses for the origin of Chinese cabbage: domesticated directly from
pak choi ​(Song ​et al.​ 1990; Zhao ​et al.​ 2005; Takuno ​et al.​ 2006)​ or via genetic admixture of
turnip and pak choi​(Li 1981; Ren ​et al.​ 1995)​. Although many contemporary studies have
included large and diverse samples, the major limitation to distinguishing ​B. rapa​ subspecies and
crops may be the number and quality of the genetic markers. Fine scale resolution of
relationships among groups within ​B. rapa​ may be beyond the power of traditional marker
systems because of their relatively short evolutionary history with high rates of outcrossing.
5
bioRxiv preprint first posted online Mar. 20, 2017; doi: http://dx.doi.org/10.1101/118372. The copyright holder for this preprint (which was not
peer-reviewed) is the author/funder. It is made available under a CC-BY 4.0 International license.
Clarification of the number of genetically distinct groups and their relationships is needed to
address basic questions about the domestication of ​B. rapa​.
Genome assemblies of multiple ​Brassica​ species ​(Wang ​et al.​ 2011; Liu ​et al.​ 2014;
Parkin ​et al.​ 2014; Chalhoub ​et al.​ 2014)​ and advancements in sequencing technology provide
an opportunity to resolve the relationships among subspecies and the domestication history of ​B.
rapa​. Recently, Cheng et al. ​(2016)​ resequenced 199 ​B. rapa​ accessions. Their STRUCTURE
and phylogenomic analyses provided improved support for the relationships among the primary
subspecies within ​B. rapa​. However, their sampling did not include several important Eurasian ​B.
rapa​ crops that are crucial for understanding the timing and order of ​B. rapa​ domestication. For
example, no samples were collected from Central Asia, the potential center of origin for ​B. rapa​.
Cheng et al. (2016) also only sampled one yellow sarson accession and did not include brown
sarson (subsp. ​dichotoma​) or rapini (subsp.​ sylvestris​). Diverse sampling of accessions
combined with population genomic approaches are needed to better resolve the complex series
of events that occurred during domestication of the​ B. rapa ​crops.
In this study, we sequenced the transcriptomes of 126 accessions representative of the
geographic and crop diversity of ​B. rapa.​ We analyzed thousands of SNPs from across the
nuclear genome to resolve the genetic structure and relationships of ​B. rapa​ crops. Using a
diffusion approximation approach, ​∂a∂i ​(Gutenkunst ​et al.​ 2009)​, we evaluated models of the
demographic history and diversification of ​B. rapa​, as well as estimated the timing of
domestication events in China and India. These age estimates were compared to the written
records available for each phase of ​B. rapa​ domestication. We also estimated the reduction of
genetic diversity and effective population size in each Asian ​B. rapa​ subspecies due to the
potential founder effect during their introduction and domestication in Asia. Our analyses provide
6
bioRxiv preprint first posted online Mar. 20, 2017; doi: http://dx.doi.org/10.1101/118372. The copyright holder for this preprint (which was not
peer-reviewed) is the author/funder. It is made available under a CC-BY 4.0 International license.
new insights into the origins and relationships of ​B. rapa​ crops. Given that most studies of
domestication are focused on crops that arose during prehistory, our analyses of ​B. rapa ​provide
a unique opportunity to explore crop domestication and diversification with a written record.
Materials and Methods
Plant Materials
We collected seeds of 143 ​B. rapa​ accessions, with 139 samples from USDA GRIN
database (​http://www.ars-grin.gov/​) and four from the authors’ collection. Samples were selected
to maximize representation of major ​B. rapa​ subspecies. These collections include​ B. rapa​ and
its 11 subspecies from 20 countries worldwide (Table S1). Six of the major subspecies have
more than eight representatives, randomly chosen from different sources, with the exception of
the two ​sylvestris ​lines collected by Dr. Gómez Campo. The Asian vegetable subspecies
parachinensis​, ​narinosa​, ​nipposinica​ and ​perviridis​ are likely derived from either or both
chinensis​ and ​pekinensis​ (Cheng et al. 2016). Therefore, only representative accessions of
these Asian vegetable subspecies were included in this analysis. After we removed misidentified
seed accessions (see below), 126 ​B. rapa ​accessions remained for the following analyses.
Plant Growth Conditions and Tissue Collection
All seeds were grown in a Conviron growth chamber with 16 hours of light at 23 ℃. and 8
hours of dark at 20 ℃. at the Bond Life Science Center, University of Missouri (Columbia,
Missouri, USA) during March to May 2015. All samples were taken from the second youngest
leaf 20 days after germination, then snap-frozen in liquid nitrogen and stored at -80 ℃.
7
bioRxiv preprint first posted online Mar. 20, 2017; doi: http://dx.doi.org/10.1101/118372. The copyright holder for this preprint (which was not
peer-reviewed) is the author/funder. It is made available under a CC-BY 4.0 International license.
Sample Preparation, RNA Isolation, Library Construction
Total RNA was isolated using the PureLinkTM RNA Mini Kit (Invitrogen, USA), following
the instructions. RNA samples were qualified and quantified by a Nanodrop 1000
spectrophotometer (Nanodrop Technologies, USA). After that, double-stranded cDNA were
synthesized according to the manufacturer’s protocol of the Maxima H Minus Double-Stranded
cDNA Synthesis Kit (Thermo, Lithuania) and purified by the GeneJET PCR Purification Kit
(Thermo, Lithuania). All samples were sequenced with Illumina HiSeq 2000 system at the
University of Missouri DNA core. Each sample contained about 9 million 250 bp pair-end reads.
SRA files for the samples are available at NCBI-SRA (SRP072186,
http://www.ncbi.nlm.nih.gov/sra/SRP072186​).
Sequence Assembly and Alignment
We first trimmed each fastq file using Trimmomatic ​(Bolger ​et al.​ 2014)​. Read quality was
assessed with FASTQC ​(Andrews & Others 2010)​. The reference ​B. rapa​ genome (version 1.5)
was downloaded from the Brassica Database (​http://brassicadb.org/brad/​).
For the reference based assemblies, we used Tophat version 2.0.14 ​(Trapnell ​et al.
2009)​ with Bowtie2 version 2.2.5 ​(Langmead & Salzberg 2012)​. Variant calling was performed
using Samtools ​(Li ​et al.​ 2009)​. We retained only SNPs and did not consider indels or MNPs in
subsequent analyses. To obtain high quality SNPs for downstream analyses, we performed
multiple SNP filtering steps: the SNPs were first filtered with ​vcfutils.pl​ script in ​SAMtools/​bcftools
package ​(Li ​et al.​ 2009)​ and ​vcffilter​ to only include SNPs with depth greater than 10 and
mapping quality greater than 30. For the phylogenetic and genetic structure analyses, we then
filtered the remaining SNPs with PLINK version 1.9 ​(Purcell ​et al.​ 2007)​ to only include SNPs
8
bioRxiv preprint first posted online Mar. 20, 2017; doi: http://dx.doi.org/10.1101/118372. The copyright holder for this preprint (which was not
peer-reviewed) is the author/funder. It is made available under a CC-BY 4.0 International license.
with a > 90% genotyping rate and minor allele frequency (MAF) higher than 5%. In contrast, our
filtering for the ​∂a∂i ​demographic models and nucleotide diversity analyses retained many
more SNPs to estimate the allele frequency spectrum. Following filtering for depth and mapping
quality, we annotated the SNPs with SnpEff ​(Cingolani ​et al.​ 2012)​ and only retained
synonymous SNPs. Further, we implemented a 25 kb sliding window neutrality test (Tajima’s D)
in vcftools ​(Danecek ​et al.​ 2011) to filter out SNPs that may be directly or indirectly under
selection.
Identification of Misidentified ​Brassica​ Seed Contaminants
Misidentified seed accessions are not uncommon in plant germplasm collections and can
confound analyses if not recognized. To identify potential seed contamination from ​B. napus​ and
B. oleracea​, we BLASTed the ​de novo​ assembled reading frames of each newly sequenced
accession against the latest ​B. rapa, B. oleracea​ (TO1000 version 2.1; genomevolution.org) and
B. napus​ (version 4.1; brassicadb.org) genomes. ​De novo​ assemblies for each accession were
performed in SOAPdenovo-Trans version 1.03 ​(Xie ​et al.​ 2014)​ with a 127 kmer. Accessions
with a majority of best BLAST hits to ​B. oleracea​ and/or ​B. napus​ were flagged as potential
contaminants. We further verified the suspicious accessions through DNA C-value measurement
using flow cytometry. Fresh leaf samples were sent to the Flow Cytometry Lab at the Benaroya
Research Institute (WA) for flow cytometry. The C-values of ​B. rapa​, ​B. oleraceae​, and ​B. napus
are significantly different from each other and provide an alternative method of verification. A
total of 11 suspicious ​B. napus,​ two suspicious ​B. oleracea​ and four suspicious autopolyploid ​B.
rapa ​were identified among the ​B. rapa​ accessions from the USDA germplasm seed bank (Table
S2). These accessions were removed from further analyses in this study. After contaminant
9
bioRxiv preprint first posted online Mar. 20, 2017; doi: http://dx.doi.org/10.1101/118372. The copyright holder for this preprint (which was not
peer-reviewed) is the author/funder. It is made available under a CC-BY 4.0 International license.
removal, 126 ​B. rapa​ accessions remained for the following analyses.
Phylogenetic Inference and Genetic Structure
SNPs called from mapping reads to the ​Brassica rapa ​reference genome were used for
phylogenetic inference and genetic structure analyses. The SNP dataset was concatenated from
ped format to fasta format using a custom perl script. The data set included SNPs from our 126
B. rapa​ accessions as well as two ​B. oleracea​ as outgroups (SRA accession: SRR1032050 and
SRR630924, http://www.ncbi.nlm.nih.gov/sra).​ ​A maximum likelihood phylogeny was constructed
using RAxML version 8.2.4 ​(Stamatakis 2014)​. We implemented RAxML’s rapid bootstrap
algorithm with the GTR+GAMMA model and rooted the phylogeny using the two ​B. oleracea
accessions. Genetic structure among the accessions was inferred using a Bayesian clustering
method implemented in fastStrucutre ​(Raj ​et al.​ 2014)​. We tested ​K​ = 1 to 10 clusters with five
replicates for each ​K ​using the default convergence criterion​ ​and prior. The optimal ​K​ value was
estimated with the ​chooseK​ tool contained in the fastStructure package. The results of all
replicates were summarized using CLUMPAK (http://clumpak.tau.ac.il/)​(Kopelman ​et al.​ 2015)​.
The multiple ​K​ value fastStructure results were re-ordered and visualized in Excel based on the
sample order of the phylogenetic inference.
Inference of Demographic History with ∂a∂i
Models of the demographic history of the major ​B. rapa​ groups were evaluated using
diffusion approximations to the allele frequency spectrum (AFS) with ​∂a∂i ​(Gutenkunst ​et al.
2009)​. This approach allowed us to rapidly and flexibly test alternative demographic models.
Only individuals that could be readily attributed to one of the major genetic lineages identified in
10
bioRxiv preprint first posted online Mar. 20, 2017; doi: http://dx.doi.org/10.1101/118372. The copyright holder for this preprint (which was not
peer-reviewed) is the author/funder. It is made available under a CC-BY 4.0 International license.
our fastStructure and phylogenetic analyses were used in this analysis (denoted by the vertical
bars on Figure 1). Individuals were assigned a group if they had >50% fastStructure identity and
were phylogenetically clustered. We analyzed the demographic history of 102 ​B. rapa
accessions divided into five groups (Table S1): 22 accessions of European and Central Asian ​B.
rapa ​(group “EU-CA”), 7 accessions representing ​B. rapa​ subsp. ​sylvestris​ (group “S”), 20
accessions of ​B. rapa​ subsp. ​trilocularis​ and ​B. rapa​ subsp. ​dichotoma ​(group “I”), 25 accessions
representing ​B. rapa​ subsp. ​chinensis ​(group “C”), and 28 accessions of ​B. rapa​ subsp.
pekinensis ​(group “P”). Based on the phylogeny, we also combined groups S and I as a South
Asian group “SA” in some analyses. Likewise, groups C and P were merged in some analyses
as an East Asian group “EA”.
We implemented a model testing hierarchy to accommodate ∂a∂i’s limit of analyzing
three populations at a time. We first evaluated one- and two-population models for each of the
five ​B. rapa​ groups to estimate the timing of divergence among the derived groups. The results
of these initial analyses informed our subsequent demographic modeling of three different
three-population model groups (Fig. 2; Table 1) described in further detail below. Notably, we
also performed the same simulations under exponential growth models, but the models did not
converge (results not shown).
Model Group A. ​In model group A, we tested whether the SA and EA groups were independently
derived from EU-CA, or whether a group ancestral to both SA and EA split from EU-CA and then
split into SA and EA (Fig. 2a).
Model Group B. ​Model group B focused on the origins of S and I. We tested whether S and I
11
bioRxiv preprint first posted online Mar. 20, 2017; doi: http://dx.doi.org/10.1101/118372. The copyright holder for this preprint (which was not
peer-reviewed) is the author/funder. It is made available under a CC-BY 4.0 International license.
independently split from EU-CA, or if a common ancestor of S and I split from EU-CA and then
subsequently diverged into S and I (Fig. 2b).
Model Group C. ​The models in group C evaluated long debated alternative hypotheses on the
origin of Chinese cabbage,​ B. rapa​ subsp. ​pekinensis​ (Fig. 2c). Especially as to whether it was
originated from ​chinensis​ ​(Song ​et al.​ 1990; Zhao ​et al.​ 2005; Takuno ​et al.​ 2006)​, from EU-CA
rapa, or from the admixture of the first two ​(Li 1981; Ren ​et al.​ 1995)​.
In each model, the EU-CA population was assumed to have constant effective population
size. Each of these models was parameterized by the timing between each split and admixture
event (e.g. T1, T2) and the relative effective population sizes over those time intervals (e.g. ​ν​EA,
ν​SA). The effective population sizes were relative to the ancestral population size, so ​ν​ less than
one indicates a population size decline, while ​ν​ greater than one indicates growth. Under the
third ​pekinensis​ origin model (C v P 3; Fig. 2C), an additional parameter f (between 0 and 1)
estimated the fraction of the admixed ​pekinensis ​population from EU-CA.
Model Optimization and Testing.​ Our filtered SNP collections were converted to the ​∂a∂i
frequency spectrum file format using a publicly available perl script
(https://github.com/owensgl/reformat/blob/master/vcf2dadi.pl). In order to correct for linkage
among the SNPs, we performed a likelihood ratio test on the models for each population using
the Godambe Information Matrix (GIM)(Godambe, 1960). This test statistic is compared to a χ​2
distribution with degrees of freedom equal to the difference in number of parameters between
the simple and complex model. The linkage of SNPs requires us to adjust the test statistic, which
12
bioRxiv preprint first posted online Mar. 20, 2017; doi: http://dx.doi.org/10.1101/118372. The copyright holder for this preprint (which was not
peer-reviewed) is the author/funder. It is made available under a CC-BY 4.0 International license.
is calculated via the GIM. We constructed the three-population folded joint frequency spectra by
projecting down to a specified sample size for each population ​(Marth ​et al.​ 2004)​. Because
many SNPs were not called in every sequenced individual, by projecting the frequency spectrum
to a smaller sample size, we were able to include more SNPs in our analysis. The sample size
that we projected to in each population was chosen so that the frequency spectrum maximized
the total number of segregating sites for each population. We projected to 20, 40, and 40
samples for the EU-CA, SA, and EA populations in the model group A analysis, respectively. For
model group B, we projected to 13, 40, and 14 samples for the EU-CA, ​trilocularis​, and ​sylvestris
populations, resp. For model group C, we projected to 11, 13, and 23 for the EU-CA, ​chinensis​,
and ​pekinensis​ populations. For ​B. rapa​ subsp. ​sylvestris​, ​trilocularis​, and ​dichotoma​, many
SNPs for which genotypes were missing from some individuals showed a large excess of
heterozygotes, and this pattern was not seen for SNPs called in every individual. Thus, we used
SNPs called in every individual in the ​sylvestris​, ​trilocularis​, and ​dichotoma​ group to avoid any
artifacts in the projected frequency spectrum.
During optimization we imposed lower and upper bounds on model parameters, with
times allowed between 0.001 and 1 genetic time units, and population size changes between
0.001 and 100 of the reference (ancestral) population size. We used built in maximum likelihood
optimization in ​∂a∂i to fit the expected frequency spectra under the​ parameterized models to
the observed spectra. The population scaled mutation rate θ was a free parameter of the
models. We repeated the optimization for each model 100 times from random initial parameter
guesses, to ensure consistent convergence to the optimal parameters. The Godambe
bootstrapping method ​(Coffman ​et al.​ 2016)​ was performed to estimate the 95% confidence
intervals of the best fit parameters. Bootstrap datasets were generated with a custom python
13
bioRxiv preprint first posted online Mar. 20, 2017; doi: http://dx.doi.org/10.1101/118372. The copyright holder for this preprint (which was not
peer-reviewed) is the author/funder. It is made available under a CC-BY 4.0 International license.
script.
We selected the best fitting model for each group of models by selecting the model with
highest log-likelihood under the best-fit parameter set. For the best fitting models in each group,
we converted the inferred population genetic parameters to years using the relation θ = 4​Neμ​L
and T = 2​Ne​g. L is the effective length in base pairs of the region from which the SNP data was
obtained. L was calculated as the product of mean gene length, total genes covered by the SNP
dataset, and the fraction of SNP data set used in the simulation due to missing data. Upper and
lower mutation rates of ​μ​ = 1.5 × 10​-8​ and 9 × 10​-9​ per synonymous site per generation and
generation time of one year were used in our age estimates ​(Koch ​et al.​ 2000; Kagale ​et al.
2014)​.
Genetic Diversity Analysis
To assess the level of genetic diversity of the five identified genetic groups, we estimated
the average nucleotide diversity, π ​(Nei & Li 1979)​, using vcftools version 0.1.12 ​(Danecek ​et al.
2011)​. A window size of 100 kbp and step size of 25 kbp were applied to the genetic diversity
estimation. All calculations and visualizations were made in R (version 3.2.2) and Rstudio
(Version 0.99.482).
Results
Read Mapping and SNP Filtering
Using paired-end Illumina RNA-seq, we sequenced the transcriptomes of 143 B
​ rassica
rapa​ accessions with 250 bp mean read length. Following the removal of 17 misidentified
14
bioRxiv preprint first posted online Mar. 20, 2017; doi: http://dx.doi.org/10.1101/118372. The copyright holder for this preprint (which was not
peer-reviewed) is the author/funder. It is made available under a CC-BY 4.0 International license.
accessions were not clearly ​B. rapa​ (Table S2), the transcriptomes of 126​ B. rapa​ accessions
were used in our analyses. On average, 8.09 million reads per sample were left after
Trimmomatic read cleaning. For each of the reference genome based assemblies, 33.90% (±
8.51%) of the reads mapped to the ​B. rapa​ genome on average. In total, 2.05 million SNPs and
0.11 million indels were identified, with 13,760 multi-allelic SNP sites, average SNP depth 4.69
(± 1.97), and a transition/transversion ratio of 1.34. After filtering, 31,662 high quality SNPs were
retained for our phylogenetic and genetic structure analyses and 426,740 SNPs were retained
for our demographic analyses. A summary of SNPs on each chromosome is available in Table
S3.
Inference of Genetic Structure and Phylogenetic Relationships
Our genetic structure and phylogenetic analyses of 31,662 SNPs supported five distinct
genetic clusters among the 126 ​B. rapa ​accessions. Analyses of genetic structure recovered six
clusters using fastStructure (Fig. 1, Fig. S1). Increasing or decreasing the number of clusters by
one did not substantially change the observed clustering pattern (Fig. 1, Fig. S1), and ​K​ = 6
clusters maximized the marginal likelihood (Fig. S2). Most of the morphologically distinct crops
long recognized as subspecies—pak choi, Chinese cabbage, rapini, and the yellow
sarson—were largely resolved as distinct clusters in our fastStructure analyses. Notably, they
are also well resolved with ​K​ = 5 or 7 clusters (Fig. 1). We also identified a number of genetically
diverse accessions (multi-color in fastStructure results, Fig. 1). Some of these accessions are
recently admixed crop lines from the US, but the history of many of these accessions is unclear.
The sixth cluster (gray color on Fig. 1) resolved in the fastStructure ​K​ = 6 analysis described
genetic variation in these diverse accessions. Otherwise, fastStructure resolved five largely
15
bioRxiv preprint first posted online Mar. 20, 2017; doi: http://dx.doi.org/10.1101/118372. The copyright holder for this preprint (which was not
peer-reviewed) is the author/funder. It is made available under a CC-BY 4.0 International license.
homogenous clusters (Fig. 1). Analyses with CLUMPAK found that the same genotypes were
consistently assigned to these clusters across ten iterations of fastStructure (Fig. S3).
The phylogenetic relationships of the ​B. rapa​ accessions were well resolved and
consistent with our fastStructure clustering. More than 98% of the nodes on the RAxML tree had
bootstrap support higher than 50% (Fig. S2). Many of the morphologically distinct crops were
distinguished as well supported clades in our phylogenomic analysis. The combined
phylogenetic and genetic clustering analyses resolved five major genetic groups within ​B. rapa
that largely corresponded to different crops. These groups were a European-Central Asian
cluster (group “EU-CA”) represented by ​B. rapa​, a group that contained rapini and other
accessions (group “S”) represented by ​B. rapa​ subsp. ​sylvestris​ (bootstrap support 100%), an
Indian sarson group (group “I”) comprised of ​B. rapa​ subsp. ​trilocularis ​and three accessions of
subsp. ​dichotoma ​(bootstrap support of 100%), a chinensis group (group “C”) composed of ​B.
rapa ​subsp.​ chinensis ​(bootstrap support of 68%), and most of the ​B. rapa​ subsp. ​pekinensis
accessions clustered together in a ​pekinensis​ group (group “P”; bootstrap support of 66%).
Interestingly, the eight ​B. rapa​ subsp. ​oleifera ​(blue stars on Fig. 1 and Fig. S2), six ​B. rapa
subsp. ​rapa​ (pink stars on Fig.1 and Fig. S2)​ ​and five of the ​dichotoma​ accessions (green stars
on Fig. 1 and Fig. S2) were distributed across the phylogeny. Genetically heterogeneous
accessions identified in our fastStructure analyses were scattered across the phylogeny. Groups
of these heterogenous samples were sister to each of the five largely homogenous genetic
groups. Whether these heterogeneous accessions are the result of recent admixture or
represent ancestral variation is not clear from the current analyses.
The phylogenetic analyses also supported a European-Central Asian origin for ​B. rapa.​
Using two ​B. oleracea ​accessions as an outgroup, the concatenated phylogenetic analysis of
16
bioRxiv preprint first posted online Mar. 20, 2017; doi: http://dx.doi.org/10.1101/118372. The copyright holder for this preprint (which was not
peer-reviewed) is the author/funder. It is made available under a CC-BY 4.0 International license.
31,662 SNPs resolved the European-Central Asian​ B. rapa​ cluster (EU-CA) as sister to all other
B. rapa​ accessions (Fig. 1; Fig. S2). Included in this clade were most of the European and
Central Asian accessions we sampled (​B. rapa​). This relationship was well supported with a
bootstrap value of 71%. The nested set of sister relationships among the remaining taxa
indicated an eastward expansion of ​B. rapa​ crops. Accessions from East and South Asia were
among the most derived lineages in our phylogeny.
Demographic History of ​Brassica rapa’​ s Domestication Across Asia
Our ∂a∂i analyses of the demographic history of ​B. rapa​ also supported an eastward
series of domestication events over the past few thousand years (Fig. 3 and Table 1). The
demographic models were well supported qualitatively: the chosen optimal project size
maximized the number of alleles captured in SFS under each model (Fig. S4); and the best fit
model for each group reproduced the observed frequency spectra with no severe deviations in
the residuals (Fig. 4). Across the range of ​B. rapa​, our data supported a stepwise eastward
progression of crop domestication over the past 2446–4070 years (Fig. 3b and Table 1) and
were best explained by a model that involved an initial population diverging from EU-CA,
followed by SA split with EA (Fig. 2a, SA v EA 3; Fig. 3a). This model was supported with a
much higher log-likelihood than the next best fitting models (Δ =2469; log-likelihood = -16390).
Under this model, we inferred the time span from initial divergence of SA-EA to the split of
SA-EA was approximately 976–1630 years, and the divergence of EA and SA occurred
approximately 1470–2440 years ago (Table 1, Fig. 3).
Among the models exploring the relationships of the primarily European, Central Asian,
and South Asian accessions (Fig. 2b), our analyses supported an initial common ancestor split
17
bioRxiv preprint first posted online Mar. 20, 2017; doi: http://dx.doi.org/10.1101/118372. The copyright holder for this preprint (which was not
peer-reviewed) is the author/funder. It is made available under a CC-BY 4.0 International license.
from EU-CA followed by subsequent divergence into the ​sylvestris​ (“S”) and ​trilocularis​ +
dichotoma​ (“I”) groups (S v I 3). The log-likelihood value of this model (-4959) is much higher
than that of the other two alternative models (Δ =12280). Under this optimal model, a population
split from EU-CA approximately 3715–6190 years ago (Table 1). This population subsequently
split into groups S and I (T1 under S v I 3 in Table 1 and Fig. 2) 825–1380 years later. This time
span before divergence of later groups is consistent with the corresponding estimate for the
duration of a common ancestral population in the optimal model of model group A (SA vs EA 3;
976–1630 years). Under this model, we inferred the split of S and I occurred 2890–4810 years
ago.
Our simulations on the origins of Chinese cabbage (​B. rapa​ subsp. ​pekinensis​) supported
an ancient admixed origin of the crop. The optimal demographic model under indicated that
pekinensis was formed by admixture of ​chinensis ​and​ rapa ​(model C v P 3, Fig. 2c). This optimal
model had a log-likelihood value (-18110) much higher than that in the two alternative models (Δ
= 383). An estimated 83% of the ​pekinensis ​genetic variation came from ​chinensis​, whereas the
remaining 17% was derived from European-Central Asian ​rapa. ​Admixture that contributed to the
founding of ​pekinensis​ is estimated to be ancient. The best fitting demographic model (Table 1)
estimated the admixture occurred 1244–2070 (± 52–86) years ago.
Our hierarchical model testing in ∂a∂i also allowed us to compare the relative effective
population sizes (​Ne​) of the ​B. rapa​ genetic groups. Given ​Ne​EU-CA​ ​was the reference standard
for the derived populations under each model,​ ​and the estimated ​Ne v​ alues​ ​for each EU-CA
-8 ​
group were similar across the three model groups (​Ne​EU-CA =
​ 22700–25900 if μ​1​ = 1.5 × 10​ or
Ne​EU-CA​ = 37800–43100 if μ​2​ = 9 × 10​-9​; Table 1), it is reasonable to compare the estimated ​Ne
for each group across models. Under the “S v I 3” and “C v P 3”​ ∂a∂i analyses​, we observed
18
bioRxiv preprint first posted online Mar. 20, 2017; doi: http://dx.doi.org/10.1101/118372. The copyright holder for this preprint (which was not
peer-reviewed) is the author/funder. It is made available under a CC-BY 4.0 International license.
four to seven times larger ​Ne​ in ​chinensis ​(​Ne​C​ = ​7760–12900) and ​pekinensis ​(​Ne​P​ =
4660–7760) than that in ​sylvestris ​(​Ne​S​ = 1
​ 100–1830) and ​trilocularis​ and ​dichotoma ​(​Ne​I​ =
1190–1990). This pattern is consistent with the results based on model “EA v SA 3” under which
Ne​EA​ ​is approximately 13 times larger than​ Ne​SA​.
Nucleotide Diversity of the ​B. rapa​ Genetic Groups
Finally, we also compared the genetic diversity of the five genetic groups of ​B. rapa
identified in our analyses. The mean nucleotide diversity of the ​sylvestris​ group (π​S​ = 1.08 × 10​-5​)
and Indian ​trilocularis​ and ​dichotoma​ ​sarson group (π​I​ = 0.97 × 10​-5​) were 63% and 66% lower
than EU-CA group (π​EU-CA​ = 2.88 × 10​-5​). In contrast, the ​chinensis​ (π​C​ = 2.45 × 10​-5​) and
pekinensis​ (π​P​ = 2.40 × 10​-5​) groups did not have a similarly strong decrease of nucleotide
diversity (15% and 17% lower) relative to the EU-CA group (Fig. 5).​ Together, our ∂a∂i
analyses ​and nucleotide diversity analyses suggested a less severe genetic bottleneck with at
least one ancient admixture event in the East Asian group relative to the South Asian group
during the eastward progression of ​B. rapa​.
Discussion
Our analyses of more than 2 million SNPs across 126 newly sequenced genotypes of
Brassica rapa​ clarified the history and origins of multiple ​Brassica​ crops. We found strong
evidence for at least five genetic groups within ​B. rapa​ from our combined phylogenetic, genetic
structure, and demographic analyses. Most previous studies identified two or three
geographically based groups: Europe-Central Asia and East Asia ​(Zhao ​et al.​ 2005; Takuno ​et
al.​ 2006)​, and South Asia if sampled ​(Song e
​ t al.​ 1990; Del Carpio ​et al.​ 2011a; b; Guo ​et al.
19
bioRxiv preprint first posted online Mar. 20, 2017; doi: http://dx.doi.org/10.1101/118372. The copyright holder for this preprint (which was not
peer-reviewed) is the author/funder. It is made available under a CC-BY 4.0 International license.
2014)​. Our results are largely consistent with these analyses. However, we found more
resolution of genetic structure among East Asian crops and recognized a genetically distinct
cluster that contains ​B. rapa​ subsp. ​sylvestris​. These results are also consistent with a recent
genomic analysis that used ~20,000 SNPs to study ​B. rapa​ relationships (Cheng et al. 2016).
Although our sampling differs from Cheng et al. (2016)—we sampled rapini (subsp. ​sylvestris​)
and brown sarson (subsp. ​dichotoma​), more yellow sarson (subsp. ​trilocularis​), and more
diverse European and Central Asian accessions—both of our analyses recovered Chinese
cabbage, pak choi, and a European-Central Asian group that includes European turnips. Their
sampling included more Asian crops than our analyses, and they resolved additional groups that
included caixin. (subsp. ​parachinenesis​), zicatai (subsp. ​chinensis​ ​var. purpurea​), and Japanese
turnips with mizuna (subsp. ​nipposinica)​ . In contrast, we sampled many accessions not
previously analyzed with genomic data, including rapini (subsp. ​sylvestris​) and the Indian
sarsons (subsp. ​dichotoma​ and ​trilocularis​), and recovered distinct genetic groups associated
with these accessions. Further, our more sophisticated modeling of the demographic history of
B. rapa​—corroborated by written records—provides new insights on the age and relationships
among the diverse ​B. rapa​ lineages. Considering the age estimates of population splits among
the ​B. rapa​ groups, it is not surprising that past analyses with fewer markers were unable to
resolve the genetic structure among these plants. Although our analyses provide strong
evidence for at least five groups, not all individuals of ​B. rapa​ could be placed in one of these
distinct genetic clusters. Determining whether this is due to parallel evolution, admixture (recent
or ancient), insufficient sampling of related genotypes, or lack of divergence requires further
research with new samples. Regardless, our analyses, as well as those of others ​(Song ​et al.
1990; Zhao ​et al.​ 2005; Takuno ​et al.​ 2006; Del Carpio ​et al.​ 2011a; b; Guo ​et al.​ 2014; Cheng ​et
20
bioRxiv preprint first posted online Mar. 20, 2017; doi: http://dx.doi.org/10.1101/118372. The copyright holder for this preprint (which was not
peer-reviewed) is the author/funder. It is made available under a CC-BY 4.0 International license.
al.​ 2016)​, contrast with one recent study ​(Tanhuanpää ​et al.​ 2016)​ and find evidence that
geography and crop type contribute to the genetic structure of ​B. rapa​.
One of the interesting aspects of the domestication of​ B. rapa​ crops is that many were
domesticated during written history. This provided an opportunity to compare our population
genomic results to historical records for many of the demographic splits modeled with ∂a∂i.
Previous simulations of ∂a∂i’s accuracy found that inferences of divergence times had low
error with even modest sample sizes ​(Robinson ​et al.​ 2014)​. A ​recent empirical analysis using
∂a∂i recovered evidence of known recent demographic changes in the checkerspot butterfly
(McCoy ​et al.​ 2014)​. Our inferred times of demographic events in the history of ​B. rapa​ are
consistent with these previous analyses and appear to be accurate as they were consistent with
the available written record. Demographic inferences based on genome wide SNP data found
that the eastward introduction and diversification of ​B. rapa​ happened 2400–4100 years ago.
This inferred time span is compatible with the two oldest historical records of ​B. rapa ​in East and
South Asia.​ Brassica ​(Li 1981; Ye 1989; Luo 1992)​ was first described in the ancient Aryan
literature of India nearly 3500 years ago ​(Prakash ​et al.​ 2011)​, and in Shi Jing (Classic of
Poetry), the oldest existing Chinese classical poetry collection ​(Li 1981; Ye 1989; Luo 1992)​.
Similarly, our ∂a∂i analyses estimated Chinese cabbage (​B. rapa​ subsp. ​pekinensis​) was
developed 1200–2100 years ago. The earliest historical record of Chinese cabbage is nearly
1400 years old. Niu Du Song (i.e., beef tripe cabbage) is mentioned in Xin Xiu Ben Cao (Newly
Revised Canon of Materia Medica), one of the oldest official pharmacy books edited by Jing Su
in 659 CE. A clear description of Chinese cabbage is also found in Song Su’s Ben Cao Tu Jing
(Illustrated Classics of Materia Medica) from 1061 CE ​(Li 1981)​. He described Niu Du Song in
Yangzhou city in Southern China as a plant with large, round, crinkled leaves and a less fibrous
21
bioRxiv preprint first posted online Mar. 20, 2017; doi: http://dx.doi.org/10.1101/118372. The copyright holder for this preprint (which was not
peer-reviewed) is the author/funder. It is made available under a CC-BY 4.0 International license.
texture. Beyond the timing of these events, the inferred order of domestication is also supported
by the written records. Our results suggests that turnips are likely the earliest domesticated crop
with leaf and seed crops domesticated later. A translated passage in the Shi Jing supports this
order: when collecting turnip and radish, we should not abandon them them because of their
bitter root because the leaves and stems taste good. These written records corroborate our
population genomic inference of the timing of two different events in the history of ​B. rapa
domestication and lend confidence to the other estimated parameters.
Notably, the inferred timing of these demographic events better overlapped with
estimates from one of the Brassicaceae substitution rates. Substitution rates in the Brassicaceae
have recently been debated as a new fossil provided a different calibration for the family
(Beilstein ​et al.​ 2010)​. The placement of this fossil has been the subject of debate ​(Franzke ​et al.
2016)​ because phylogenetic inferences generally yield older dates for macroevolutionary events
(Beilstein ​et al.​ 2010; Arias ​et al.​ 2014)​ than previous estimates ​(Koch ​et al.​ 2000; Barker ​et al.
2009; Edger ​et al.​ 2015; Hohmann ​et al.​ 2015)​. To work within this debate, our demographic
estimates were calculated twice using two different estimates of the Brassicaceae nuclear
substitution rate: the faster rate from Kock et al. (2000) and a more recent, slower rate
developed by Kagale et al. (2014) based on the fossil placement of Beilstein et al. (2010). Three
of the four written records occurred within the 95% confidence interval for the timing of
population splits estimated with μ​1​, whereas all of the estimates using μ​2​ preceded the available
written records. Thus, both of these substitution rates are consistent with the written record
although there is greater concordance with the more recently derived and slightly slower
substitution rate. Considering that we expect a written record to follow a domestication event
with some lag, the “true” substitution rate for ​Brassica​ and other Brassicaceae may lay
22
bioRxiv preprint first posted online Mar. 20, 2017; doi: http://dx.doi.org/10.1101/118372. The copyright holder for this preprint (which was not
peer-reviewed) is the author/funder. It is made available under a CC-BY 4.0 International license.
somewhere between these two rate estimates.
The ∂a∂i modeling also indicated that the South Asian group (“S” and “I”) experienced
a more severe founder effect than the East Asian group (“EA”). ​The estimated effective
population size and nucleotide diversity of the East Asian group were several-fold larger than the
South Asian group. This difference of founder effect within a single species is striking in
comparison to other crops. The South Asian group retains only approximately 35% of the genetic
diversity of the EU-CA group. This is equivalent to the domestication bottleneck in barley
(Caldwell ​et al.​ 2006)​, rice ​(Zhu ​et al.​ 2007; Caicedo ​et al.​ 2007)​, and among the most severe
bottlenecks found in annual crops ​(Miller & Gross 2011)​. In contrast, the East Asian group
retained approximately 84% of the genetic diversity present in the EU-CA group. The
domestication bottleneck of the East Asian ​B. rapa​ crops is comparable to the bottleneck in
sorghum (​ Casa ​et al.​ 2005)​, sunflower ​(Tang & Knapp 2003)​, soybean ​(Li ​et al.​ 2010)​, and
among the least severe bottlenecks found in annual crops ​(Miller & Gross 2011)​. Considering
the two lineages diverged only a few thousand years ago and both likely experienced
long-distance introductions from ancestral European-Central Asian populations, the difference of
Ne ​and π between the two groups is probably due to a combination of differences in gene flow
(Meyer & Purugganan 2013)​ and inbreeding ​(Dempewolf ​et al.​ 2012)​. In East Asia, commercial
activity along the ancient silk-road (started 220 BC) may have continuously brought diverse B
​ .
rapa ​genetic resources to East Asia. Consistent with this hypothesis, we found evidence of
shared genetic variation among some European-Central Asian and East Asian accessions, as
well as a small amount of EU-CA admixture in the origin of​ B. rapa​ subsp. ​pekinensis​. In
contrast, South Asian ​B. rapa​ may have been more punctually introduced with events such as
Alexander the Great’s invasion of India in 325 BCE. Written records document his troops likely
23
bioRxiv preprint first posted online Mar. 20, 2017; doi: http://dx.doi.org/10.1101/118372. The copyright holder for this preprint (which was not
peer-reviewed) is the author/funder. It is made available under a CC-BY 4.0 International license.
brought mustard seeds as army provisions ​(Grubben 2004)​. A shift to predominant inbreeding in
the yellow and brown sarsons ​(McGrath & Quiros 1992)​ likely further reduced their genetic
diversity. Thus, domestication of these two different regional groups of ​B. rapa​ crops occurred
with significantly different population genetic backgrounds that essentially spans the range of
possible models of domestication ​(Eyre-Walker ​et al.​ 1998; Miller & Gross 2011; Meyer &
Purugganan 2013; Gaut ​et al.​ 2015)​. This less severe founder effect in the EA group may
contribute to the diverse crop types in East Asia, in contrast to the Indian group.
Despite the economic importance of ​B. rapa​, the order of crop domestication has never
been well resolved. Previous studies generally considered ​B. rapa ​subsp. ​sylvestris​ to be the
wild type and reflect a European​ B. rapa​ center of origin ​(Guo ​et al.​ 2014; Tanhuanpää e
​ t al.
2016)​. However, ​B. rapa​ subsp. ​sylvestris​ has never been genetically distinguished from other ​B.
rapa​ subspecies. Genome-wide analyses with 715 polymorphic SSR alleles ​(Guo ​et al.​ 2014)
and 209 nuclear SNPs ​(Tanhuanpää ​et al.​ 2016)​ grouped ​B. rapa​ subsp. ​sylvestris​ among
European accessions or clustered with oil types. A recent whole genome resequencing study of
B. rapa​ (Cheng et al 2016) did not include any ​sylvestris​ accessions. Our results suggest
sylvestris​ is​ ​not likely a wild ​B. rapa.​ The two ​sylvestris​ accessions and its five genetically
cognate accessions in our analyses clustered into a monophyletic clade that was supported with
100% bootstrap support in our phylogenetic analysis. These accessions were also grouped in
our genetic structure analyses and present in multiple values of ​K​. Importantly, this group was
not sister to all of the remaining ​B. rapa​ accessions. Using two ​B. oleracea​ as outgroups to root
the phylogenetic tree of ​B. rapa​, we found a collection of European and Central Asian ​B. rapa
accessions were sister to the remaining ​B. rapa​. Based on recent whole genome resequencing
analyses ​(Cheng ​et al.​ 2016)​, previous linguistic evidence ​(Ignatov ​et al.​ 2008)​, and
24
bioRxiv preprint first posted online Mar. 20, 2017; doi: http://dx.doi.org/10.1101/118372. The copyright holder for this preprint (which was not
peer-reviewed) is the author/funder. It is made available under a CC-BY 4.0 International license.
archaeological records ​(Körber-Grohne 1987)​, turnip is likely the first domesticated ​Brassica​ in
the European-Central Asian region. Our study is less clear on this question as only one of the
turnip accessions was placed in the EU-CA clade. However, none of our turnip collections are
from the European-Central Asian region, and ⅔ of them have a complex genetic background
based on our fastStructure analyses. Our result is congruent with recent reports by Cheng et al.
(2016)​ that there may be two turnip lineages (European and Asian turnip). Further sampling of
turnips from across the geographic range would address many of the remaining questions
surrounding the early domestication of ​B. rapa.​
With an expanded sample of yellow sarson and brown sarson, we also found evidence
for a single origin of yellow sarson (​B. rapa​ subsp. ​trilocularis​), but possibly multiple origins of
brown sarson (​B. rapa ​subsp. ​dichotoma​). All accessions of ​trilocularis​ and three of the
dichotoma​ accessions formed a monophyletic clade in our phylogenetic analysis. This is
consistent with a single origin of the yellow sarsons with multiple origins of brown sarsons from
yellow sarsons. Further, a couple individuals of ​dichotoma​ clustered more closely with ​sylvestris
and a single individual was grouped with ​chinensis​. Other ​dichotoma​ individuals were distributed
among the genetically heterogeneous accessions. Assuming that our accessions are correctly
labeled, these results suggest that brown sarsons may have more than one origin from
non-sarson ancestors. It is worth noting that the USDA nomenclature system groups brown
sarson together with another crop, toria, as subsp. ​dichotoma​. This may be the cause of
apparent polyphyly of subsp. ​dichotoma​. Alternatively, the placement of some ​dichotoma
accessions may reflect more recent admixture rather than independent origins. These results are
mostly consistent with a previous analysis based on five isozyme and four RFLP markers ​(1992)​.
Other similar ​B. rapa​ population and phylogenetic studies typically adopted or mentioned these
25
bioRxiv preprint first posted online Mar. 20, 2017; doi: http://dx.doi.org/10.1101/118372. The copyright holder for this preprint (which was not
peer-reviewed) is the author/funder. It is made available under a CC-BY 4.0 International license.
two subspecies, but only included a few samples to draw phylogenetic inferences (e.g., Cheng ​et
al. ​2016 and ​Del Carpio ​et al.​ 2011b​). Thus, our results indicate that the potential diversity and
relationships of these morphologically similar crops have yet to be fully explored. Additional
sampling beyond the available USDA accessions of the yellow and brown sarsons—including
both self-compatible and incompatible types as well as clearly identified representatives of
toria—is needed to further resolve the origins and diversity of these crops.
Oil crops are among the most economically important crops of ​B. rapa​, and we found
evidence for multiple origins of oil type ​B. rapa​ subsp. ​oleifera​. The eight ​oleifera s​ amples were
distributed throughout our phylogeny and occurred in four of the five major genetic groups.
Considering that by definition canola is a ​Brassica​ species containing less than 2% erucic acid
and less than 30 micromoles glucosinolates ​(Eskin & McDonald 1991)​, it is not surprising to find
oleifera​ samples scattered across the phylogeny. Oil crops are recognized as being genetically
and geographically diverse relative to other cultivated crops because of likely multiple origins of
oilseed crops ​(McGrath & Quiros 1992; Takuno ​et al.​ 2006; Tanhuanpää ​et al.​ 2016)​. Our results
also support parallel origins of the oil types of ​B. rapa​ from different genetic backgrounds. Oil
type ​B. rapa​ appears to have been primarily selected from mostly non-leafy crops with a
concentration among the European-Central Asian genetic group.The oil crops of ​Brassica​ may
have multiple, independent origins. Studies on oilseed rape (​Brassica napus​) have also shown
that they have relatively diverse accessions ​that have been domesticated for winter and spring
oilseed rape in Europe and Asia​(Allender & King 2010; Li ​et al.​ 2014; Wu ​et al.​ 2014; Gazave ​et
al.​ 2016)​. Considering that most ​Brassica​ oil cultivars have been domesticated relatively recent
(Reiner ​et al.​ 1995)​ that post-date the globalization of agriculture ​(Phillips & Khachatourians
2001; Font ​et al.​ 2003; Kumar ​et al.​ 2015)​, it is not surprising that they were sampled from
26
bioRxiv preprint first posted online Mar. 20, 2017; doi: http://dx.doi.org/10.1101/118372. The copyright holder for this preprint (which was not
peer-reviewed) is the author/funder. It is made available under a CC-BY 4.0 International license.
throughout the phylogeny. Thus, the domestication of oil type ​B. rapa ​may similar to the parallel
domestication of those oil types in ​B.napus​.
Our relatively large sampling and demographic analyses allowed us to resolve a
long-held debate on the origin of Chinese cabbage. Previous genetic and empirical studies
supported two distinct hypotheses for the origin of Chinese cabbage: directly from pak choi
(Song ​et al.​ 1990; Zhao ​et al.​ 2005; Takuno ​et al.​ 2006)​ or via genetic admixture of turnip and
pak choi ​(Li 1981; Ren ​et al.​ 1995)​. Our fastStructure and phylogenetic inference suggest
pekinensis ​has a closer relationship with ​chinensis. ​Our demographic modeling supported an
admixed origin of ​pekinensis​ via crossing of EU-CA ​Brassica rapa​ and subsp. ​chinensis, ​with
chinensis​ contributing ~90% of the contemporary ​pekinensis​ genome​. ​Given that most of the
genome is ​chinensis​, it is not unexpected that previous analyses with fewer markers would not
detect the modest 10% genomic contribution from European-Central Asian ​B. rapa​. The
fastStructure results may also indicate the ​B. rapa​ subsp. ​rapa​ accessions most closely related
to the progenitor of ​pekinensis​. Components of genetic variation in a handful of ​B. rapa​ subsp.
rapa​ accessions were assigned to the same category as the bulk of our ​pekinensis​ genotypes. It
is not clear from the fastStructure analysis if this structure represents the ancestral variation
contributed by ​rapa​ to ​pekinensis​, or more recent admixture from ​pekinensis​ into some ​rapa
accessions. Detailed analyses and expanded sampling are required to further dissect the
contributions and history of ​B. rapa ​subsp. ​pekinensis​. Our analyses also demonstrate that
Chinese cabbage was formed much later than other Asian ​B. rapa​ crops and indicate it is among
the most recently derived ​B. rapa​ subspecies. Samples of other East Asian leafy type
subspecies (​parachinensis​, ​narinosa, perviridis ​and ​nipposinica​) were placed in either the
pekinensis​ or ​chinensis​ clades, consistent with a recent study with abundant sampling of these
27
bioRxiv preprint first posted online Mar. 20, 2017; doi: http://dx.doi.org/10.1101/118372. The copyright holder for this preprint (which was not
peer-reviewed) is the author/funder. It is made available under a CC-BY 4.0 International license.
leafy types ​(Cheng ​et al.​ 2016)​.
Conclusions
Overall, our study provides many new data and insights into the history of ​B. rapa​ and
provides clear direction for future research. Our expanded sampling of frequently
underrepresented groups, such as the Indian sarsons, has provided evidence on their origins
and diversity. Expanded sampling of genetic diversity within ​B. rapa ​will likely resolve additional
groups and clarify the complex history of these plants. In particular, better genomic sampling of
B. rapa ​landraces from Central Asia, landrace turnip from Europe, landraces and elite rapini from
Italy, landrace toria, and oil type ​oleifera​ are needed to further resolve the domestication history
of ​B. rapa​. These samples would augment the existing collections of ​B. rapa​ in North America
and aid curation of the existing USDA collection which contained many misidentified accessions.
Sequence data from wild ​B. rapa ​seeds or archeological materials would also significantly
improve our analyses. Additional genomes from putative wild​ B. rapa ​populations would also
mitigate ascertainment bias in future analyses and provide more insight into changes in
structural variation associated with domestication. Despite potential for further improvement, our
results have provided the most robust evidence to date on the genetic structure and
domestication history of ​B. rapa​. Our use of demographic modeling provided some of the first
population genomic estimates of the ages and population size changes that occurred during​ B.
rapa​’s history. Corroboration of the estimated dates with the written record provides added
confidence in our inferences. Given that the other ​Brassica​ crops are of similar age with equally
complicated histories, the population genomic approach employed here will likely provide new
insights into ​Brassica​ domestication when applied to the other crops. Most importantly, our
28
bioRxiv preprint first posted online Mar. 20, 2017; doi: http://dx.doi.org/10.1101/118372. The copyright holder for this preprint (which was not
peer-reviewed) is the author/funder. It is made available under a CC-BY 4.0 International license.
results provide a new framework for ​B. rapa​ that includes empirical estimates of domestication
times, effective population sizes, and the amount and direction of gene flow between divergent
lineages.
Acknowledgements:
We thank A. Baniaga, K. Dlugosch, S. Jorgensen, H. Marx, and L. Zheng for helpful comments
on the manuscript. Server hosting infrastructure and services provided by the Bio Computing
Facility (BCF) at the University of Arizona. X.Q., T.E.H., H.A., J.C.P., and M.S.B. were supported
by NSF-IOS-1339156. A.R. and R.G. were supported by NSF-DEB-1146074.
References
Allender CJ, King GJ (2010) Origins of the amphiploid species Brassica napus L. investigated by
chloroplast and nuclear molecular markers. ​BMC plant biology​, ​10​, 54.
Andrews S, Others (2010) FastQC: A quality control tool for high throughput sequence data.
Reference Source​.
Arias T, Beilstein MA, Tang M, McKain MR, Pires JC (2014) Diversification times among
Brassica (Brassicaceae) crops suggest hybrid formation after 20 million years of divergence.
American journal of botany​, ​101​, 86–91.
Barker MS, Vogel H, Schranz ME (2009) Paleopolyploidy in the Brassicales: analyses of the
Cleome transcriptome elucidate the history of genome duplications in Arabidopsis and other
Brassicales. ​Genome biology and evolution​, ​1​, 391–399.
Beilstein MA, Nagalingum NS, Clements MD, Manchester SR, Mathews S (2010) Dated
molecular phylogenies indicate a Miocene origin for Arabidopsis thaliana. ​Proceedings of
the National Academy of Sciences​, ​107​, 18724–18728.
Bolger AM, Lohse M, Usadel B (2014) Trimmomatic: a flexible trimmer for Illumina sequence
data. ​Bioinformatics ​, ​30​, 2114–2120.
Caicedo AL, Williamson SH, Hernandez RD ​et al.​ (2007) Genome-wide patterns of nucleotide
polymorphism in domesticated rice. ​PLoS genetics,​ ​3​, 1745–1756.
Caldwell KS, Russell J, Langridge P, Powell W (2006) Extreme population-dependent linkage
disequilibrium detected in an inbreeding plant species, Hordeum vulgare. ​Genetics​, ​172​,
557–567.
Casa AM, Mitchell SE, Hamblin MT ​et al.​ (2005) Diversity and selection in sorghum:
simultaneous analyses using simple sequence repeats. ​Theoretical and applied genetics​,
111​, 23–30.
Chalhoub B, Denoeud F, Liu S ​et al.​ (2014) Early allopolyploid evolution in the post-Neolithic
Brassica napus oilseed genome. ​Science​, ​345​, 950–953.
Cheng F, Sun R, Hou X ​et al.​ (2016) Subgenome parallel selection is associated with
29
bioRxiv preprint first posted online Mar. 20, 2017; doi: http://dx.doi.org/10.1101/118372. The copyright holder for this preprint (which was not
peer-reviewed) is the author/funder. It is made available under a CC-BY 4.0 International license.
morphotype diversification and convergent crop domestication in Brassica rapa and
Brassica oleracea. ​Nature genetics​, ​48​, 1218–1224.
Cingolani P, Platts A, Wang LL ​et al.​ (2012) A program for annotating and predicting the effects
of single nucleotide polymorphisms, SnpEff: SNPs in the genome of Drosophila
melanogaster strain w1118; iso-2; iso-3. ​Fly​, ​6​, 80–92.
Coffman AJ, Hsieh PH, Gravel S, Gutenkunst RN (2016) Computationally efficient composite
likelihood statistics for demographic inference. ​Molecular biology and evolution​, ​33​,
591–593.
Danecek P, Auton A, Abecasis G ​et al.​ (2011) The variant call format and VCFtools.
Bioinformatics ​, ​27​, 2156–2158.
Darwin C (1859) ​On the origin of the species by natural selection​. Murray.
Del Carpio DP, Basnet RK, De Vos RCH ​et al.​ (2011a) The patterns of population differentiation
in a Brassica rapa core collection. ​Theoretical and applied genetics​, ​122​, 1105–1118.
Del Carpio DP, Basnet RK, De Vos RCH ​et al.​ (2011b) Comparative methods for association
studies: a case study on metabolite variation in a Brassica rapa core collection. ​PloS one​, ​6​,
e19624.
Dempewolf H, Hodgins KA, Rummell SE, Ellstrand NC, Rieseberg LH (2012) Reproductive
isolation during domestication. ​The Plant cell​, ​24​, 2710–2717.
Diamond J (2002) Evolution, consequences and future of plant and animal domestication.
Nature​, ​418​, 700–707.
Edger PP, Heidel-Fischer HM, Bekaert M ​et al.​ (2015) The butterfly plant arms-race escalated by
gene and genome duplications. ​Proceedings of the National Academy of Sciences of the
United States of America​, ​112​, 8362–8366.
Eskin NAM, McDonald BE (1991) Canola Oil. ​Nutrition bulletin​, ​16​, 138–146.
Eyre-Walker A, Gaut RL, Hilton H, Feldman DL, Gaut BS (1998) Investigation of the bottleneck
leading to the domestication of maize. ​Proceedings of the National Academy of Sciences of
the United States of America​, ​95​, 4441–4446.
Font R, del Rio M, Fernández JM, de Haro A (2003) Acid detergent fiber analysis in oilseed
Brassicas by near-infrared spectroscopy. ​Journal of agricultural and food chemistry​, ​51​,
2917–2922.
Franzke A, Koch MA, Mummenhoff K (2016) Turnip time travels: age estimates in Brassicaceae.
Trends in plant science​.
Fuller DQ, Qin L, Zheng Y ​et al.​ (2009) The domestication process and domestication rate in
rice: spikelet bases from the Lower Yangtze. ​Science​, ​323​, 1607–1610.
Gaut BS, Díez CM, Morrell PL (2015) Genomics and the contrasting dynamics of annual and
perennial domestication. ​Trends in genetics​, ​31​, 709–719.
Gazave E, Tassone EE, Ilut DC ​et al.​ (2016) Population genomic analysis reveals differential
evolutionary histories and patterns of diversity across subgenomes and subpopulations of
Brassica napus L. ​Frontiers in plant science​, ​7​, 525.
Godambe VP (1960) An Optimum Property of Regular Maximum Likelihood Estimation. ​Annals
of Mathematical Statistics​, ​31​, 1208–1211.
Gross BL, Olsen KM (2010) Genetic perspectives on crop domestication. ​Trends in plant
science​, ​15​, 529–537.
Gross BL, Zhao Z (2014) Archaeological and genetic insights into the origins of domesticated
rice. ​Proceedings of the National Academy of Sciences of the United States of America​,
111​, 6190–6197.
30
bioRxiv preprint first posted online Mar. 20, 2017; doi: http://dx.doi.org/10.1101/118372. The copyright holder for this preprint (which was not
peer-reviewed) is the author/funder. It is made available under a CC-BY 4.0 International license.
Grubben GJH (2004) ​Vegetables​. Prota.
Guo Y, Chen S, Li Z, Cowling WA (2014) Center of origin and centers of diversity in an ancient
crop, Brassica rapa (Turnip rape). ​The Journal of heredity,​ ​ esu021​.
Gutenkunst RN, Hernandez RD, Williamson SH, Bustamante CD (2009) Inferring the joint
demographic history of multiple populations from multidimensional SNP frequency data.
PLoS genetics​, ​5​, e1000695.
Hohmann N, Wolf EM, Lysak MA, Koch MA (2015) A time-calibrated road map of Brassicaceae
species radiation and evolutionary history. ​The Plant cell​, ​27​, 2770–2784.
Ignatov AN, Artemyeva AM, Hida K (2008) Origin and expansion of cultivated Brassica rapa in
Eurasia: linguistic facts. In:. V International Symposium on Brassicas and XVI International
Crucifer Genetics Workshop, Brassica 2008 867.
Kagale S, Robinson SJ, Nixon J ​et al.​ (2014) Polyploid evolution of the Brassicaceae during the
Cenozoic era. ​The Plant cell​, ​26​, 2777–2791.
Koch MA, Haubold B, Mitchell-Olds T (2000) Comparative evolutionary analysis of chalcone
synthase and alcohol dehydrogenase loci in Arabidopsis, Arabis, and related genera
(Brassicaceae). ​Molecular biology and evolution​, ​17​, 1483–1498.
Kopelman NM, Mayzel J, Jakobsson M, Rosenberg NA, Mayrose I (2015) Clumpak: a program
for identifying clustering modes and packaging population structure inferences across K.
Molecular ecology resources​, ​15​, 1179–1191.
Körber-Grohne U (1987) Nutzpflanzen in Deutschland: Kulturgeschichte und Biologie. Theiss,
Konrad.
Kovach MJ, Sweeney MT, McCouch SR (2007) New insights into the history of rice
domestication. ​Trends in genetics​, ​23​, 578–587.
Kumar A, Banga SS, Meena PD, Kumar PR (2015) ​Brassica oilseeds: breeding and
management​. CABI.
Ladizinsky G (2012) ​Plant evolution under domestication​. Springer Netherlands.
Langmead B, Salzberg SL (2012) Fast gapped-read alignment with Bowtie 2. N
​ ature methods​,
9​, 357–359.
Li CW (1981) The origin, evolution, taxonomy and hybridization of chinese cabbage. In: ​Chinese
cabbage. Proceedings of the first international symposium.​, pp. 3–10. AVRDC.
Li F, Chen B, Xu K ​et al.​ (2014) Genome-wide association study dissects the genetic
architecture of seed weight and seed quality in rapeseed (Brassica napus L.). ​DNA
research​, ​21​, 355–367.
Li H, Handsaker B, Wysoker A ​et al.​ (2009) The sequence alignment/map format and SAMtools.
Bioinformatics ​, ​25​, 2078–2079.
Li Y-H, Li W, Zhang C ​et al.​ (2010) Genetic diversity in domesticated soybean (Glycine max) and
its wild progenitor (Glycine soja) for simple sequence repeat and single-nucleotide
polymorphism loci. ​The New phytologist​, ​188​, 242–253.
Liu S, Liu Y, Yang X ​et al.​ (2014) The Brassica oleracea genome reveals the asymmetrical
evolution of polyploid genomes. ​Nature communications​, ​5​, 3930.
Long A, Benz BF, Donahue DJ, Jull AJT, Toolin LJ (2006) First direct AMS dates on early maize
from Tehuacan, Mexico. ​Radiocarbon​, ​31​, 1035–1040.
Luo G (1992) A analysis of historical materials about the time and place of the origin of the
peking cabbage. ​Studies in the history of natural sciences​, ​2​, 171–176.
Marth GT, Czabarka E, Murvai J, Sherry ST (2004) The allele frequency spectrum in
genome-wide human variation data reveals signals of differential demographic history in
31
bioRxiv preprint first posted online Mar. 20, 2017; doi: http://dx.doi.org/10.1101/118372. The copyright holder for this preprint (which was not
peer-reviewed) is the author/funder. It is made available under a CC-BY 4.0 International license.
three large world populations. ​Genetics​, ​166​, 351–372.
McCoy RC, Garud NR, Kelley JL, Boggs CL, Petrov DA (2014) Genomic inference accurately
predicts the timing and severity of a recent bottleneck in a nonmodel insect population.
Molecular ecology​, ​23​, 136–150.
McGrath JM, Quiros CF (1992) Genetic diversity at isozyme and RFLP loci in Brassica
campestris as related to crop type and geographical origin. ​Theoretical and applied
genetics​, ​83​, 783–790.
Meyer RS, Purugganan MD (2013) Evolution of crop species: genetics of domestication and
diversification. ​Nature reviews genetics​, ​14​, 840–852.
Miller AJ, Gross BL (2011) From forest to field: perennial fruit crop domestication. ​American
journal of botany​, ​98​, 1389–1414.
Molina J, Sikora M, Garud N ​et al.​ (2011) Molecular evidence for a single evolutionary origin of
domesticated rice. ​Proceedings of the National Academy of Sciences of the United States of
America​, ​108​, 8351–8356.
Nei M, Li WH (1979) Mathematical model for studying genetic variation in terms of restriction
endonucleases. ​Proceedings of the National Academy of Sciences of the United States of
America​, ​76​, 5269–5273.
Olsen KM, Wendel JF (2013) A bountiful harvest: genomic insights into crop domestication
phenotypes. ​Annual review of plant biology​, ​64​, 47–70.
Pang W, Li X, Choi SR ​et al.​ (2015) Development of a leafy Brassica rapa fixed line collection for
genetic diversity and population structure analysis. ​Molecular breeding: new strategies in
plant improvement​, ​35​, 1–15.
Parkin IAP, Koh C, Tang H ​et al.​ (2014) Transcriptome and methylome profiling reveals relics of
genome dominance in the mesopolyploid Brassica oleracea. ​Genome biology​, ​15​, R77.
Phillips PWB, Khachatourians GG (2001) ​The biotechnology revolution in global agriculture:
innovation, invention, and investment in the Canola industry​. CABI Pub.
Piperno DR, Flannery KV (2001) The earliest archaeological maize (Zea mays L.) from highland
Mexico: new accelerator mass spectrometry dates and their implications. ​Proceedings of the
National Academy of Sciences of the United States of America​, ​98​, 2101–2103.
Piperno DR, Ranere AJ, Holst I, Iriarte J, Dickau R (2009) Starch grain and phytolith evidence
for early ninth millennium BP maize from the Central Balsas River Valley, Mexico.
Proceedings of the National Academy of Sciences​, ​106​, 5019–5024.
Prakash S, Wu X-M, Bhat SR (2011) History, evolution, and domestication of Brassica crops. In:
Plant Breeding Reviews​, pp. 19–84. Wiley Online Library.
Purcell S, Neale B, Todd-Brown K ​et al.​ (2007) PLINK: a tool set for whole-genome association
and population-based linkage analyses. ​American journal of human genetics​, ​81​, 559–575.
Purugganan MD, Boyles AL, Suddith JI (2000) Variation and selection at the CAULIFLOWER
floral homeotic gene accompanying the evolution of domesticated Brassica oleracea.
Genetics​, ​155​, 855–862.
Purugganan MD, Fuller DQ (2009) The nature of selection during plant domestication. ​Nature​,
457​, 843–848.
Raj A, Stephens M, Pritchard JK (2014) fastSTRUCTURE: variational inference of population
structure in large SNP data sets. ​Genetics,​ ​197​, 573–589.
Reiner H, Holzner W, Ebermann R (1995) The development of turnip-type and oilseed-type
Brassica rapa crops from the wild-type in Europe--An overview of botanical, historical and
linguistic facts. ​Rapeseed Today Tomorrow​, ​4​, 1066–1069.
32
bioRxiv preprint first posted online Mar. 20, 2017; doi: http://dx.doi.org/10.1101/118372. The copyright holder for this preprint (which was not
peer-reviewed) is the author/funder. It is made available under a CC-BY 4.0 International license.
Ren J, McFerson JR, Li R, Kresovich S, Lamboy WF (1995) Identities and relationships among
Chinese vegetable Brassicas as determined by Random Amplified Polymorphic DNA
markers. ​Journal of the American Society for Horticultural Science. American Society for
Horticultural Science​, ​120​, 548–555.
Robinson JD, Coffman AJ, Hickerson MJ, Gutenkunst RN (2014) Sampling strategies for
frequency spectrum-based population genomic inference. B
​ MC evolutionary biology​, ​14​,
254.
Ross-Ibarra J, Morrell PL, Gaut BS (2007) Plant domestication, a unique opportunity to identify
the genetic basis of adaptation. ​Proceedings of the National Academy of Sciences of the
United States of America​, ​104​, 8641–8648.
Sang T, Ge S (2013) Understanding rice domestication and implications for cultivar
improvement. ​Current opinion in plant biology​, ​16​, 139–146.
Smith BD (1997) The initial domestication of Cucurbita pepo in the Americas 10,000 years ago.
Science​, ​276​, 932–934.
Song K, Osborn TC, Williams PH (1990) Brassica taxonomy based on nuclear restriction
fragment length polymorphisms (RFLPs). ​Theoretical and applied genetics​, ​79​, 497–506.
Stamatakis A (2014) RAxML version 8: a tool for phylogenetic analysis and post-analysis of
large phylogenies. ​Bioinformatics ​, ​30​, 1312–1313.
Takahashi Y, Yokoi S, Takahata Y (2015) Genetic divergence of turnip (Brassica rapa L. em.
Metzg. subsp. rapa) inferred from simple sequence repeats in chloroplast and nuclear
genomes and morphology. ​Genetic resources and crop evolution​, ​63​, 869–879.
Takuno S, Kawahara T, Ohnishi O (2006) Phylogenetic relationships among cultivated types of
Brassica rapa L. em. Metzg. as revealed by AFLP analysis. ​Genetic resources and crop
evolution​, ​54​, 279–285.
Tang S, Knapp SJ (2003) Microsatellites uncover extraordinary diversity in native American land
races and wild populations of cultivated sunflower. ​Theoretical and applied genetics​, ​106​,
990–1003.
Tanhuanpää P, Erkkilä M, Tenhola-Roininen T, Tanskanen J, Manninen O (2016) SNP diversity
within and among Brassica rapa accessions reveals no geographic differentiation. ​Genome​,
59​, 11–21.
Trapnell C, Pachter L, Salzberg SL (2009) TopHat: discovering splice junctions with RNA-Seq.
Bioinformatics ​, ​25​, 1105–1111.
Wang X, Wang H, Wang J ​et al.​ (2011) The genome of the mesopolyploid crop species Brassica
rapa. ​Nature genetics​, ​43​, 1035–1039.
Wu J, Li F, Xu K ​et al.​ (2014) Assessing and broadening genetic diversity of a rapeseed
germplasm collection. ​Breeding science​, ​64​, 321–330.
Xie Y, Wu G, Tang J ​et al.​ (2014) SOAPdenovo-Trans: de novo transcriptome assembly with
short RNA-Seq reads. ​Bioinformatics ​, ​30​, 1660–1666.
Ye J (1989) The name and reality of the rape (Brassica campestris and Brassica juncea) in
China. ​Studies in the history of natural sciences​, ​8​, 158–165.
Zhao J, Artemyeva A, Del Carpio DP ​et al.​ (2010) Design of a Brassica rapa core collection for
association mapping studies. ​Genome,​ ​53​, 884–898.
Zhao J, Paulo M-J, Jamar D ​et al.​ (2007) Association mapping of leaf traits, flowering time, and
phytate content in Brassica rapa. ​Genome​, ​50​, 963–973.
Zhao J, Wang X, Deng B ​et al.​ (2005) Genetic relationships within Brassica rapa as inferred from
AFLP fingerprints. ​Theoretical and applied genetics​, ​110​, 1301–1314.
33
bioRxiv preprint first posted online Mar. 20, 2017; doi: http://dx.doi.org/10.1101/118372. The copyright holder for this preprint (which was not
peer-reviewed) is the author/funder. It is made available under a CC-BY 4.0 International license.
Zhu Q, Zheng X, Luo J, Gaut BS, Ge S (2007) Multilocus analysis of nucleotide variation of
Oryza sativa and its wild relatives: severe bottleneck during domestication of rice. ​Molecular
biology and evolution​, ​24​, 875–888.
Zohary D, Hopf M, Weiss E (2012) ​Domestication of plants in the Old World: the origin and
spread of domesticated plants in Southwest Asia, Europe, and the Mediterranean basin​.
Oxford University Press.
34
bioRxiv preprint first posted online Mar. 20, 2017; doi: http://dx.doi.org/10.1101/118372. The copyright holder for this preprint (which was not
peer-reviewed) is the author/funder. It is made available under a CC-BY 4.0 International license.
Data Accessibility:
SRA files for the samples are available at NCBI-SRA database (SRP072186,
http://www.ncbi.nlm.nih.gov/sra/SRP072186).
Author Contributions:
M.S.B., J.C.P. and R.G. designed the experiments. X.Q., H.A., A.R. and T.E.H performed the
experiments, X.Q. and M.S.B. wrote the paper.
35
bioRxiv preprint first posted online Mar. 20, 2017; doi: http://dx.doi.org/10.1101/118372. The copyright holder for this preprint (which was not
peer-reviewed) is the author/funder. It is made available under a CC-BY 4.0 International license.
Table 1.​ The parameters and confidence intervals inferred in ∂a∂i simulations. These
parameters correspond to those displayed in Fig 3. The 95% CI was calculated using the
Godambe bootstrapping method. Unit of absolute effective population size: individual; unit of
absolute time: year.
Model
Parameter
Relative Value
(​ν​)
95%CI
Absolute Value
95%CI
(μ1 = 1.5E-8)
Absolute Value
95%CI
(μ2 = 9E-9)
EA v SA 3Ne​SA
0.022
0.003
501
±68
836
±114
Ne​EA
0.28
0.1
6,350
±2268
10,600
±3786
T1
0.022
0.005
976
±222
1,630
±370
T2
0.032
0.004
1,470
±184
2,440
±305
theta
8,357
log-likelihood
-16,390
Ne​EU-CA
SvI3
22,700
Ne​I
NeS​
0.048
0.052
0.004
0.004
1,109
1,206
±92
±93
1,848
2,010
±154
±155
T1
0.018
0.002
825
±93
1,380
±153
T2
0.063
0.005
2,890
±229
4,810
±382
theta
5,019
log-likelihood
-4959
Ne​EU-CA
CvP3
37,800
22,900
38,200
Ne​C
Ne​P
0.300
0.18
0.02
0.005
7,760
4,660
±517
±129
12,513
8,969
±834
±249
T1
0.038
0.004
1,970
±207
3,280
±345
T2
0.024
0.001
1,244
±52
2,070
±86
f
0.170
0.002
theta
30,959
log-likelihood
-18,110
Ne​EU-CA
25,900
43,100
36
bioRxiv preprint first posted online Mar. 20, 2017; doi: http://dx.doi.org/10.1101/118372. The copyright holder for this preprint (which was not
peer-reviewed) is the author/funder. It is made available under a CC-BY 4.0 International license.
Theta=4​Ne​μL
T = 2​Ne​g
L = Number of genes * mean gene length * r; r is the ratio of SNPs with enough calls in each model.
f: the fraction of the admixed pekinensis population from EU-CA
r: the fraction of SNP data set used in the simulation
r​SAvsEA​ = 0.189; r​SvI​ = 0.111; r​CvP =
​ 0.614
Generation time = 1 year
Mean gene length = 1173 bp
Mean number of genes captured in transcriptome per sample = 27690
μ​1​ = 1.5E-8
μ​2​ = 9E-9
37
bioRxiv preprint first posted online Mar. 20, 2017; doi: http://dx.doi.org/10.1101/118372. The copyright holder for this preprint (which was not
peer-reviewed) is the author/funder. It is made available under a CC-BY 4.0 International license.
Fig. 1.​ ​The integrated maximum likelihood phylogeny and fastStructure results of the 126
Brassica rapa​ accessions analyzed here.​ The five ​B. rapa​ genetic groups are represented by
Europe-Central Asian ​B. rapa​, rapini (subsp. ​sylvestris​), yellow sarson (subsp. ​trilocularis​), pak
choi (subsp. ​chinensis​), and Chinese cabbage (subsp. ​pekinensis​). fastStructure ancestry
proportions shown from​ K ​= 5 to 7 with optimal ​K​ = 6. Blue, green, and pink stars represent ​B.
rapa​ subsp. ​oleifera​, subsp. ​dichotoma,​ and subsp. ​rapa​ accessions. Two ​B. oleracea
accessions were used to root the phylogenetic tree. Bootstrap support values of key nodes are
labeled on the tree. Attributions for ​B. rapa​ images are in Table S4.
38
bioRxiv preprint first posted online Mar. 20, 2017; doi: http://dx.doi.org/10.1101/118372. The copyright holder for this preprint (which was not
peer-reviewed) is the author/funder. It is made available under a CC-BY 4.0 International license.
Fig. 2.​ ​The demographic models evaluated with ​∂a∂i.​ The models with gray background
color are the optimal model under each scenario. The Log-likelihood value is indicated under
each model. EU-CA represents ​B. rapa​ and ​B. rapa​ subsp. ​rapa​ group; SA represents ​B. rapa
subsp. ​sylvestris​, ​B. rapa​ subsp. ​trilocularis ​and ​B. rapa ​subsp. ​dichotoma​ group; EA represents
B. rapa​ subsp. ​chinensis​ and ​B. rapa​ subsp. ​pekinensis​ group; S represents ​B. rapa​ subsp.
sylvestris​ group; I represents ​B. rapa​ subsp. ​trilocularis ​and ​B. rapa ​subsp. d
​ ichotoma​ group; C
represents ​B. rapa​ subsp. ​chinensis​ group; P represents ​B. rapa​ subsp. ​pekinensis​ group.
39
bioRxiv preprint first posted online Mar. 20, 2017; doi: http://dx.doi.org/10.1101/118372. The copyright holder for this preprint (which was not
peer-reviewed) is the author/funder. It is made available under a CC-BY 4.0 International license.
40
bioRxiv preprint first posted online Mar. 20, 2017; doi: http://dx.doi.org/10.1101/118372. The copyright holder for this preprint (which was not
peer-reviewed) is the author/funder. It is made available under a CC-BY 4.0 International license.
Fig. 3. A combined summary of demographic inferences and written records of ​B. rapa
domestication events. (a) ​T​he combined best-fitting demographic models based on ∂a∂i
simulations with absolute parameter values. ​The colored parallelograms represent different ​B.
rapa​ genetic groups: European-Central Asian ​B. rapa​ and ​B. rapa​ subsp. ​rapa​ (purple); ​B. rapa
subsp. ​sylvestris​ (orange); ​B. rapa​ subsp. ​trilocularis​ and ​B. rapa​ subsp. ​dichotoma​ (green); B
​ .
rapa​ subsp. ​pekinensis​ (yellow); ​B. rapa​ subsp. ​chinensis​ (red). The width of the branches
represents relative ​Ne​. The numbers on the parallelograms represent the effective population
size estimates (unit: individual). A horizontal band represents an admixture event with
proportions defined by parameters in the model.​ (b)​ A plot of the eastward introduction and
diversification of ​B. rapa​ with a time scale of key domestication events corroborated by historical
written records​. ​The horizontal arrow represents time scale. Gray callouts identify the historical
written records of ​Brassica rapa ​crops (cultivar, source, time, country). The purple and yellow
solid bars represent the estimated time of eastward introduction and domestication of Chinese
Cabbage in ​∂a∂i​ based on two substitution rates (μ1 and μ2), respectively. The dashed bars
represent the time range between the inferences from two different substitution rates.
41
bioRxiv preprint first posted online Mar. 20, 2017; doi: http://dx.doi.org/10.1101/118372. The copyright holder for this preprint (which was not
peer-reviewed) is the author/funder. It is made available under a CC-BY 4.0 International license.
Fig. 4.​ ​Population pairwise plots of 3D model-data comparison for each optimal model as
logarithmic color-maps under the tested models.​ (A) EA vs SA 3; (B) S vs I 3; (C) C vs P 3.
For each model, the first row is the data frequency spectrum plots, the second row is the model
frequency spectrum plots. On each heat plot, the upper-left is the model and data frequency
spectra, the lower-right is the residuals between model and data. The third and fourth row are
the plots of residuals and histograms of residuals. Dark red or blue on the residual plots indicate
the model predicts too many or few alleles. ​The residual patterns of ∂a∂i modeling ​could be
explained by ongoing gene flow between the populations, which was not included in our
demographic models.
42
bioRxiv preprint first posted online Mar. 20, 2017; doi: http://dx.doi.org/10.1101/118372. The copyright holder for this preprint (which was not
peer-reviewed) is the author/funder. It is made available under a CC-BY 4.0 International license.
Fig. 5.​ ​Box plot of nucleotide diversity of the five ​B. rapa​ genetic groups. ​Bottom and top of
the box represent the first and third quartiles. Band inside the box represent median. The
numbers represent mean values for each group. Colors of each group are consistent with
fastStructure results. Abbreviation: EU-CA represents ​B. rapa​ and ​B. rapa​ subsp. ​rapa​ group; S
represents ​B. rapa​ subsp. ​sylvestris​ group; I represents ​B. rapa​ subsp. ​trilocularis ​and ​B. rapa
subsp. ​dichotoma​ group; C represents ​B. rapa​ subsp. ​chinensis​ group; P represents ​B. rapa
subsp. ​pekinensis​ group.
43
bioRxiv preprint first posted online Mar. 20, 2017; doi: http://dx.doi.org/10.1101/118372. The copyright holder for this preprint (which was not
peer-reviewed) is the author/funder. It is made available under a CC-BY 4.0 International license.
Supplementary Tables and Figures
Table S1 Sample information of the investigated ​Brassica rapa​ accessions.
Table S2 A list of the misidentified ​Brassica​ seeds from USDA. None of these accessions were
used in this study.
Table S3 Summary of SNPs on each chromosome using ​Brassica rapa ​genomes version 1.5 as
reference.
Table S4 Image sources.
Fig. S1 Marginal likelihood distribution of the fastStructure clustering with ​K​ ranging from 1 to 10.
Python script ​chooseK​ coming with the fastStructure package was used to estimate the best ​K
value. ​K ​= 6 explained the population structure best and maximized marginal likelihood.
Fig. S2 The integrated maximum likelihood phylogeny and fastStructure results of the 126
investigated ​Brassica rapa​ accessions. An expanded version of Fig 1 with detail accession
information.
Fig. S3 A comparison of fastStructure result and CLUMPAK result based on ten iterations when
K​ = 6.
Fig S4 The trend lines illustrate the relationship between project size and the number of alleles
captured in SFS under model (a) and model (b, c).
44